demo_classification.R

source("kbmtl_semisupervised_classification_variational_train.R")
source("kbmtl_semisupervised_classification_variational_test.R")

#initalize the parameters of the algorithm
parameters <- list()

#set the hyperparameters of gamma prior used for projection matrix
parameters$alpha_lambda <- 1
parameters$beta_lambda <- 1

### IMPORTANT ###
#For gamma priors, you can experiment with three different (alpha, beta) values
#(1, 1) => default priors
#(1e-10, 1e+10) => good for obtaining sparsity
#(1e-10, 1e-10) => good for small sample size problems

#set the number of iterations
parameters$iteration <- 200

#set the margin parameter
parameters$margin <- 1

#set the subspace dimensionality
parameters$R <- 20

#set the seed for random number generator used to initalize random variables
parameters$seed <- 1606

#set the standard deviation of hidden representations
parameters$sigma_h <- 0.1

#set the standard deviation of weight parameters
parameters$sigma_w <- 1.0

#initialize the kernel and class labels for training
Ktrain <- ?? #should be an Ntra x Ntra matrix containing similarity values between training samples
Ytrain <- ?? #should be an Ntra x T matrix containing class labels (contains only -1s, +1s, and NaNs)

#perform training
state <- kbmtl_semisupervised_classification_variational_train(Ktrain, Ytrain, parameters)

#initialize the kernel for testing
Ktest <- ?? #should be an Ntra x Ntest matrix containing similarity values between training and test samples

#perform prediction
prediction <- kbmtl_semisupervised_classification_variational_test(Ktest, state)

#display the predicted probabilities
print(prediction$P)